schematch

Crates.ioschematch
lib.rsschematch
version1.0.0
sourcesrc
created_at2024-03-31 06:28:07.816814
updated_at2024-03-31 06:28:07.816814
descriptionDeclarative schema checking command
homepage
repositoryhttps://github.com/ppdx999/schematch
max_upload_size
id1191477
size56,081
fujis (ppdx999)

documentation

README

schematch

Declarative schema checking commands

Installation

cargo install schematch

Usage

Usage: schematch [OPTIONS] <SCHEMA> [FILE]

Arguments:
  <SCHEMA>  The schema to check against
  [FILE]    The file to check. If not provided, stdin will be used

Options:
  -s, --schema-type <SCHEMA_TYPE>  Schema type. schematch support tsv and json, If not provided tsv will be used [default: tsv] [possible values: tsv, json]
  -h, --help                       Print help
  -V, --version                    Print version

Quickstart

You can validate the schema of your data

for example when you have data.txt

1 john@example.com   John_Doe
2 sherry@example.com Sherry_Berry
3 ram@example.com    Ram_Singh

then you can validate your data's schema like this

$ cat data.txt | schematch "id:integer email:string name:string"
1 john@example.com   John_Doe
2 sherry@example.com Sherry_Berry
3 ram@example.com    Ram_Singh

$ echo $?
0

If your data is valid, then schematch's exit code is 0, else 1.

Schematch don't modify recieved data and sends it to stdout as it is. Therefore, schematch can collaborate flexibly with other commands through 'pipe', and can be easily integrated into existing shell script pipeliens.

ex)

$ cat data.txt                                    |
  schematch "id:integer email:string name:string" |
  awk '{print $1, $3}'                            |
  schematch "id:integer name:string"              |
  .
  .
  .

Why?

Unix shell is beautifully small and highly composable. However, they are often less readable and understandable. This is partly because it is difficult to see how each command interacts with the data. To understand exactly what a shell script is doing, you need to understand the data structure of the files read by the cat and find commands, and how awk and sed process them.

However, shell scripts only describe the latter information (= how the data is processed). The former information (= how the original data and the processed data are structured) is always in the file and is never declaratively described in the shell script.

Schematch allows the former information(= data structure) to be included in the ShellScript pipeline.

For example, to use a shell to output the number of accesses per hour from an apache log file, the following shell script might be written

cat /var/log/apache2/access.log | awk '{print $4}' | cut -b 2-15 | sort | uniq -c

This certainly works well, but it is difficult to understand which command does what.

Schematch can change this to

cat        /var/log/apache2/access.log                                       |
schematch 'ip:string localuser:string remoteuser:string time:string 
           res:string status:number bite:number referer:string agent:string' |
awk       '{print $4}'                                                       |
schematch 'time:string'                                                      |
tr        ':/' ' '                                                           |
schematch 'date:number month:string year:string h:number m:number s:number'  |
awk       '{printf "%s_%s\n", $1, $2}'                                       |
schematch 'date_month:string'                                                |
sort                                                                         |
uniq -c

It describes more declaratively what the structure of the data is in the pipeline and how each command is processing the data semantically.

Supported Schema

  • tsv
  • json

Tsv

Usage

$ cat data.txt
1 jhon_doe@example.com  true  Jhon_Doe
2 emily_lua@example.com false Emily_Lua
3 mac_kily@example.com  true  _
$ cat data.txt | schematch "id:integer  email:string  is_active:boolean  name:string|null" > /dev/null

$ echo $?
0

Supported type and value

type valid value invalid value
integer 1, 449, -4 , etc.. 1.0, x123
float 1, 4.0, -39, etc.. xx, yy
string aaa, bbb, c, etc.. String accpet anything
boolean true, false all chars other than valid case is invalid
null _ all chars other than valid case is invalid

Json

$ cat data.txt
{
  "group": "Group1",
  "members": [
    {"id": "aaa", "name": "Jhon"},
    {"id": "bbb", "name": "Mary"}
  ]
}
$ cat data.txt | schematch --schema-type json "{group: string, members: Array<{id: string, name: string}>}" > /dev/null

$ echo $?
0

Supported Types

  • null
  • string
  • number
  • boolean
  • object
Commit count: 78

cargo fmt